ontorata/ratary

What it does

Ratary is a persistent memory and knowledge management system designed specifically for AI assistants and coding workflows. It provides hybrid search capabilities, knowledge graph construction, and token-efficient context preservation, allowing AI assistants to maintain and retrieve relevant information across conversations and projects.

Key features

• Hybrid search combining semantic and keyword-based retrieval
• Knowledge graph construction for relationship mapping
• Token-efficient context management to optimize API usage
• MCP (Model Context Protocol) server with 28 tools
• Multiple database support: D1, PostgreSQL, MariaDB, or Docker
• Optional remote HTTP streaming capability
• Self-hosted deployment with full data control

How to set up

Install via npm using @ratary/mcpserver package. Configure your preferred database backend (D1, Postgres, MariaDB, or Docker). For Claude/ChatGPT integration, add the MCP server to your client configuration. Optionally enable remote HTTP streaming for distributed setups. Full documentation available on GitHub.